Colorectal Polyps Missed with Optical Colonoscopy Despite Previous Detection and Localization with CT Colonography

Nearly one in four nondiminutive colorectal polyps detected by using CT colonography but not found with subsequent nonblinded optical colonoscopy ultimately proves to be a false-negative (FN) finding, and a majority of these FN findings occur in the right colon and are clinically significant.

Abstract

Purpose

To retrospectively evaluate and characterize nondiminutive colorectal polyps prospectively detected by using computed tomographic (CT) colonography but not confirmed with subsequent nonblinded optical colonoscopy (OC).

Materials and Methods

This study was institutional review board approved; the need for signed informed consent was waived. Over 113 months, 9336 adults (mean age, 57.1 years) underwent CT colonography, which yielded 2606 nondiminutive (≥6 mm) polyps. Of 1731 polyps that underwent subsequent nonblinded OC (ie, endoscopists provided advanced knowledge of specific polyp size, location, and morphologic appearance at CT colonography), 181 (10%) were not confirmed with initial endoscopy (ie, discordant), of which 37 were excluded (awaiting or lost to follow-up). After discordant polyp review, 66 of the 144 lesions were categorized as likely CT colonography false-positive findings (no further action) and 78 were categorized as potential OC false-negative (FN) findings.

Results

Thirty-one of 144 (21.5%) of all discordant lesions were confirmed as FN findings at OC, including 40% (31 of 78) of those with OC and/or CT colonography follow-up. OC FN lesions were an average of 8.5 mm ± 3.3 in diameter and were identified with higher confidence at prospective CT colonography (on a 3-point confidence scale: mean, 2.8 vs 2.3; P = .001). OC FN findings were more likely than concordant polyps to be located in the right colon (respectively, 71% [22 of 31] vs 47% [723 of 1535]; P = .010). Most (81% [21 of 26]) OC FN lesions that were ultimately resected were neoplastic (adenomas or serrated lesions), of which 43% (nine of 21) were characterized as advanced lesions, and 89% (eight of nine) of advanced lesions occurred in the right colon.

Conclusion

In clinical practice, polyps prospectively identified with CT colonography but not confirmed with subsequent nonblinded (ie, despite a priori knowledge of the CT colonography findings) OC require additional review because a substantial proportion may be FN findings. Most FN findings found with OC demonstrated clinically significant histopathologic results, and a majority of advanced lesions occurred in the right colon.

^© RSNA, 2015

An earlier incorrect version of this article appeared online. This article was corrected on August 19, 2015.

Introduction

The miss rate of optical colonoscopy (OC) for nondiminutive colorectal polyps was previously studied by using a variety of methodologic analyses. By using back-to-back colonoscopies, Rex et al (1) estimated a per-polyp OC miss rate of 13% for small adenomas (6–9 mm) and 6% for large adenomas (≥10 mm). However, because the evaluated modality was also the reference standard, it is likely the true miss rate was underestimated in these tandem OC studies. By segmentally unblinding computed tomographic (CT) colonography results at subsequent colonoscopy, a two-fold–greater miss rate of 12% for large adenomas was demonstrated (2,3). Despite this enhanced reference standard, even these figures likely underestimate the true miss rate of OC, given that a single colonoscopy examination was used to attempt to confirm its own misses; any polyp detected by using CT colonography that was not found immediately after segmental unblinding was unavoidably classified as a CT colonography false-positive (FP) finding. We have the ability to perform longitudinal follow-up of discordant cases in clinical practice, which allows for accurate distinction between CT colonography FP findings and OC false-negative (FN) results.

The clinical importance of missed colorectal lesions should be emphasized because these lesions may ultimately progress to colorectal cancer (4). This is especially true in patients who were otherwise presumed negative for colorectal cancer and for whom screening guidelines may not recommend follow-up for up to 10 years, which could be after malignant transformation ensues. Consequently, further understanding of the nature of polyps missed with OC is critical to improve the paradigm of colorectal cancer screening.

CT colonography represents a fundamentally distinct and complementary modality for polyp detection and it is shown to be equivalent overall to OC for colorectal cancer screening in terms of detection of advanced neoplasia (5). Previous studies (6,7) suggested a wider degree of variability for lesion detection with OC compared with CT colonography. This variation in OC detection is especially pronounced for proximal serrated polyps (8), which may account in part for the relative lack of protection from right-sided colorectal cancer with OC screening (9–12). CT colonography may be complementary to OC for right-sided colonic evaluation, which allows for the possibility of increased detection of right-sided and other lesions missed by using OC. To the best of our knowledge, no studies to date focused on the ultimate reconciliation of discordant polyps labeled as CT colonography FP findings in various trials; these require additional follow-up that is lacking from validation trials. Furthermore, these potential OC misses are above and beyond the established FN results demonstrated by segmental unblinding because no blinding is performed in actual practice.

The purpose of this study is to retrospectively evaluate and characterize nondiminutive colorectal polyps prospectively detected by using CT colonography but not confirmed with subsequent nonblinded OC.

Materials and Methods

This retrospective analysis of prospectively acquired data was Health Insurance Portability and Accountability Act compliant and approved by our institutional review board. The requirement for signed informed consent was waived.

This study was performed without industrial sponsorship or support. One author (P.J.P.) previously served as a consultant for Viatronix (Stony Brook, NY) and one author (D.H.K.) remains a consultant for Viatronix, the developer of a software platform used for data interpretation, but this United States Food and Drug Administration–approved clinical product is not evaluated in our study. Data and information presented herein were under control of the primary author (B.D.P.) who has had no personal or professional relationships that may present a conflict of interest. Otherwise, the authors have no relevant disclosures.

Patient Population, Study Group, and Follow-up

A flow diagram of the study cohort is presented in Figure 1. Between April 2004 and September 2013, 9336 adult patients (mean age, 57.1 years ± 8.0 [standard deviation]; 4210 men, 5126 women) underwent a total of 10 976 screening CT colonography examinations at our single center, which yielded a total of 2606 unique nondiminutive (≥6 mm) polyps prospectively detected in 1462 patients. Included patients were generally asymptomatic adults who underwent CT colonography for colon cancer screening. A small minority were asymptomatic adults who were evaluated after incomplete OC or who returned for follow-up of previously detected colon polyps. Patients with a history of colorectal cancer, inflammatory bowel disease, polyposis syndromes, previous colorectal surgery, or with symptoms referable to the lower gastrointestinal tract were not included.

Figure 1:

Flow diagram of study cohort. *Ten of the 62 polyps with OC follow-up first underwent repeat CT colonography (CTC) to confirm the presence of a lesion.

Polyp characteristics, including lesion size (in millimeters), segmental location, morphologic appearance (sessile, flat, or pedunculated), and diagnostic confidence of the interpreting radiologist (by using three-point scale [3, most confident; 2, somewhat confident; 1, least confident]) (13) were prospectively recorded for all polyps. Flat polyps were defined as plaque-like lesions 6–29 mm in diameter and 3 mm or less in height (14). Unblinded endoscopy for polypectomy was performed in 1731 of 2606 (66.4%) unique nondiminutive polyps in 1028 patients by using either OC (92.8% [1607 of 1731]) or flexible sigmoidoscopy (4.3% [74 of 1731]). A small group of patients proceeded directly to surgery (2.9% [50 of 1731]). In cases where flexible sigmoidoscopy was used, the polyps were detected in only the sigmoid colon or rectum (the remainder of the colon was imaged and shown to be clear of polyps by using CT colonography), and the decision to proceed to flexible sigmoidoscopy was made after consulting the endoscopist and/or the patient’s primary care physician. In all cases, results of CT colonography—including polyp size, morphologic appearance, and location—were available to endoscopists and surgeons before the intervention, hence the term “nonblinded” to distinguish this from the segmental unblinding technique used in validation trials. The remaining 875 (33.6%) lesions, which consisted of small polyps that were 6–9 mm, underwent further in vivo CT colonographic surveillance (15).

Of the 1731 polyps imaged by using OC for polypectomy (or surgery) after CT colonography, we found 1550 (89.5%) concordant lesions, including all polyps and masses that proceeded surgery. The remaining 181 (10.5%) lesions were not found at the initial unblinded OC (ie, discordant lesions), despite a priori knowledge of the CT colonographic results. As part of our routine clinical practice, all discordant polyps undergo additional review by two radiologists who are CT colonography experts; neither radiologist interpreted the original CT colonography study. After this independent review, the radiologists reach a consensus regarding the likelihood of whether the discordant finding is a CT colonography FP finding or a possible OC miss (ie, a FN finding). After the review process, 66 polyps were classified as CT colonography FP findings without the necessity of further follow-up; 115 polyps were classified as possible OC misses, and additional short interval follow-up (defined as follow-up within 3 years) by using either repeat OC or repeat CT colonography was recommended. Factors that influence classification of likely CT colonography FP findings versus possible OC FN findings include predictive features, such lesion size, lesion morphologic structure, confirmation on both CT colonographic views, and location behind a fold (2,16). As of the time of this study, 17 patients (with 19 polyps) were still waiting for follow-up (these patients have not yet returned for follow-up examination) and 14 patients (with 18 polyps) were lost to follow-up. Therefore, we report on 144 of the 181 total discordant lesions, including 78 polyps that were further evaluated with repeat OC (64% [50 of 78]), CT colonography (20% [16 of 78]), flexible sigmoidoscopy (3% [two of 78]), or both OC and CT colonography (13% [10 of 78]), as directed by the endoscopist and/or the patient’s primary physician.

Operator Demographics

All CT colonography examinations were prospectively interpreted by one of 12 experienced board-certified radiologists who practices within our abdominal imaging section (mean length of time in practice, 14 years [range, 2–33 years]). This same pool of radiologists acted as independent reviewers for discordant cases, as described. Endoscopic examinations (OC or flexible sigmoidoscopy) after CT colonography were performed by board-certified gastroenterologists or surgeons within the same group practice or referral network. Over 90% of endoscopies were performed by one of 23 experienced gastroenterologists (mean length of time in practice, 18 years [range, 7–42 years]); this same pool of gastroenterologists also performed additional OC examinations in cases of potential OC misses.

CT Colonography Protocol and Technique

The CT colonography technique used at our institution was described in detail elsewhere (5,17). In summary, patients underwent a low-volume bowel preparation on the day before they underwent CT colonography by using a cathartic cleansing agent. Sodium phosphate initially was the standard cathartic, but it was later replaced by magnesium citrate (18). Oral contrast material tagging of residual fluid and fecal material was achieved with 2.1% weight per volume barium sulfate and diatrizoate. During the CT colonography examination, colonic insufflation was maintained by using automated continuous carbon dioxide delivered through a small flexible rectal catheter (19). Patients were routinely scanned in both supine and prone positions, with decubitus positioning as needed (20). Images were acquired with 8–64 section multi–detector row CT scanners by using 1.25-mm collimation, 1-mm reconstruction interval, 120 kVp, and either a fixed tube current–time product (50–75 mAs) or tube current modulation (range, 30–300 mA). Radiologist interpretation of CT colonography examinations was performed by using both three-dimensional reconstructions for initial polyp detection and two-dimensional cross-sectional images for secondary detection and polyp confirmation. All examinations were interpreted by using a dedicated CT colonography software system (Viatronix V3D Colon; Viatronix).

Statistical Analysis

Primary data analysis was principally performed by one of the authors (B.D.P.). Confirmation of an OC FN finding was made by performing subsequent OC and/or CT colonography on the patient. Although confirmation at subsequent OC allows for histologic analysis after polypectomy, confirmation by using follow-up CT colonography provides higher confidence to detect the identical lesion. When applicable, Fisher exact test and Pearson χ² test were used to test for differences in categorical variables, which included analysis of polyp morphologic structure and location. Student t test was used to test for differences in continuous variables, including polyp diameter. No adjustment for multiple comparisons was made. A two-tailed P value of less than .05 was used as the criterion to indicate statistical significance.

Results

At final follow-up assessment, 21% (31 of 144) of discordant lesions after initial nonblinded endoscopic examination were confirmed to be actual OC FN results on the basis of subsequent detection. This included 40% (31 of 78) of cases with OC and/or CT colonography follow-up evaluation. The remaining 78.5% (113 of 144) of polyps were either classified as CT colonography FP findings after independent review without the need for further follow-up (45.8% [66 of 144]) or were not found at follow-up OC or CT colonography and remained labeled as presumed CT colonography FP results (32.6% [47 of 144]). Of note, two patients each had two confirmed polyps that were OC FN findings, which brought the total to 29 of 1028 patients with OC FN findings, or a per-patient rate of OC FN findings of 2.8%. For the 78 discordant lesions that underwent further OC or CT colonography follow-up, average time to final follow-up was 2.6 years ± 1.9. For 30 of 31 (97%) ultimately confirmed polyps, the initial discordant OC examination was performed by one of the experienced group of 23 gastroenterologists responsible for over 90% of the endoscopies performed in this study; the remaining discordant polyp was missed at flexible sigmoidoscopy by a general surgeon with 13 years of experience.

The average diameter of lesions not seen by using OC was 8.5 mm ± 3.3. Compared with presumed CT colonography FP results, the morphologic structure of lesions that were missed by using OC and found at repeat testing were slightly more likely to be sessile or pedunculated, not flat (respectively, 74% [23 of 31] vs 54.9% [62 of 113]; P = .064). Lesions missed at initial OC were more likely to have higher diagnostic confidence at the prospective CT colonography examination (by using the previously discussed three-point scale) (21), with a mean diagnostic confidence of 2.8 points ± 0.5 compared with 2.2 points ± 0.7 for cases where CT colonography helped to show FP findings (P ≤ .001). Comparison of confirmed OC missed lesions with presumed CT colonography FP findings is summarized in Table 1. The lesions missed by using OC were significantly more likely to be located in the right colon compared with 1550 polyps that were concordant at the initial OC examination (respectively, 71% [22 of 31] vs 47.1% [723 of 1535]; P = .010). No significant differences were observed between discordant OC missed lesions and concordant polyps revealed by using CT colonography with OC in terms of lesion size, morphologic structure, or diagnostic confidence; these results are summarized in Table 2.

Table 1.

Comparison of Confirmed OC Misses with Presumed CT Colonography FP Findings

Note.—Unless otherwise indicated, data are percentages and data in parentheses are numerators/denominators.

*Data are means ± standard deviation.

^†Diagnostic confidence was found by using a three-point scale at CT colonography.

Table 2.

Comparison of Confirmed OC Misses with Concordant Lesions by Using CT Colonographic Imaging

Note.—Unless otherwise indicated, data are percentages with numerator and denominator in parentheses. There were 31 FN findings confirmed with OC. There were 1535 findings confirmed with concordant CT colonography and OC.

*Benign stricturing disease (n = 15) excluded.

^†Data are means ± standard deviation.

^‡Three-point scale with CT colonography (diagnostic confidence: 3, most; 2, somewhat; 1, least).

Of 26 polyps missed by using OC but confirmed by subsequent OC with polypectomy, 21 polyps (81%) were classified as potentially premalignant with pathologic analysis (one tubulovillous adenoma, 11 tubular adenomas, and nine serrated lesions). The remainder of lesions missed with OC was benign, with three hyperplastic polyps and two lesions that yielded only normal colonic mucosa with pathologic analysis. Of the potentially premalignant polyps, nine of 21 (43%) met criteria for advanced neoplasia (5,22), with eight of nine (89%) advanced neoplasms located in the right colon. One example of a lesion missed by using OC that proved to be an advanced tubular adenoma is depicted in Figure 2.

Figure 2:

Large advanced tubular adenoma missed at initial nonblinded OC after prospective detection at initial CT colonography screening. A, Three-dimensional endoluminal view with CT colonography shows a 2-cm sessile polyp (arrow) detected behind a cecal fold adjacent to the ileocecal (IC) valve. B, Coronal two-dimensional CT colonography image confirms a true mucosa-based polyp (arrow). A diminutive rectal lesion (not shown) was also detected and incidentally noted in the CT colonography report. C, D, Retroflexed images from same-day OC referral show the (C) cecum and ascending colon and (D) dedicated evaluation of the ileocecal valve fold (arrow in D). The polyp was not found despite previous knowledge of specific location and thorough inspection. The diminutive rectal lesion was confirmed and proved to be a hyperplastic polyp. The OC report recommended follow-up OC in 5 years. E, Because of the standard expert discordant review process, repeat CT colonography with same-day OC, if needed, was recommended, and repeat CT colonography 9 months later shows the 2-cm polyp (arrow). F, Repeat OC performed by a different gastroenterologist confirms the large polyp behind a fold, which was resected and proved to be a tubular adenoma, advanced according to size criteria (≥10 mm).

Discussion

Colorectal cancer remains a prevalent public health concern, and the American Cancer Society estimating over 136 000 new cases of colorectal cancer in 2014 (23). Screening for and removal of precancerous polyps has proven to be effective by decreasing both colorectal cancer incidence and mortality in the recent decades (24–26). Effective colorectal cancer screening depends on identification and resection of relevant precancerous polyps before their transformation into frank malignancy (24). Screening colonoscopy is currently recommended at 10-year intervals starting at age 50 years after an examination that is negative for polyps (27). Because of this prolonged interval, premalignant lesions missed with OC can have grave consequences for patients according to the degree of progression toward malignancy. In our study, the majority of nondiminutive lesions detected with CT colonography that were missed at the initial nonblinded OC after CT colonography proved to be premalignant with pathologic analysis, and a substantial fraction of these met the established criteria for advanced neoplasia (28). Furthermore, eight of nine advanced adenomas missed at the initial OC after CT colonography were located in the right colon, including five large serrated lesions. These lesions represent a risk to the patient, especially if there is a prolonged interval between examinations, and may help explain the phenomenon of interval cancers following so-called negative colonoscopy, as well as the relative lack of protection for right-sided colorectal cancer with OC screening (9–12).

The miss rate for OC, which was long considered the reference standard screening test for colorectal cancer, was estimated in previous studies (2) to be up to 12% for large (≥10 mm) adenomas when CT colonography results were incorporated into an enhanced referenced standard (eg, through segmental unblinding). It is well-established (2,9,11,12) that it is difficult to detect lesions both within the right colon and on the so-called back side of colonic folds at OC. Although the segmental unblinding of CT colonography results reveals a two-fold increase in OC miss rate compared with tandem colonoscopy (1), these figures still likely represent an underestimation because these previous trials lack a standard that includes longitudinal follow-up. It is important to realize that the OC FN findings in this study actually represent an additional subset of OC misses in excess of those previously reported because they were confirmed with repeat testing at a later date. OC FN findings uncovered through segmental unblinding of CT colonography results in previously reported clinical trials are not typically missed at OC in clinical practice because the CT colonography findings are not blinded. Therefore, with the benefit of advanced knowledge of CT colonography findings, the per-patient miss rate for OC in this study was 3%, which is substantially lower than that previously reported for OC alone. Although this real-world nonblinded approach with repeat studies in cases of questionable discordant lesions revealed additional OC misses, there remains the possibility of even more polyps not revealed with OC that were FN findings (although this number is likely small) because all lesions not ultimately confirmed by repeat testing are inevitably labeled as CT colonography FP findings.

CT colonography has been shown to have comparable efficacy to OC for detection of advanced neoplasia with proper technique (5), and for detection of cancer regardless of specific technique (29). Primary colorectal cancer screening with CT colonography followed by OC for polypectomy in cases with positive findings was demonstrated to be cost effective compared with primary OC screening (30–32). Current screening guidelines for CT colonography recommend a 5–10 year interval after an examination without positive findings (21), although the 5-year interval is most often quoted (33). Studies showed that a CT colonography screening study that shows no polyps that are 6 mm or greater confers protection from colorectal cancer for at least a 5-year interval, despite the fact that diminutive lesions are largely ignored (34). Our current institutional practice is to recommend repeat screening in 5 years after an initial CT colonography examination that shows no polyps 6 mm or greater and a 5–10 year interval after a second CT colonography examination that shows no polyps 6 mm or greater.

Although a 10-year interval after negative OC screening has been widely accepted, the higher rate of interval cancers reported relative to our CT colonography experience may be in large part because of relevant missed lesions (9–12,34,35). Previous studies expressed particular concern regarding right-sided lesions, a trend supported in this study: eight of nine advanced premalignant lesions in the right colon were missed with OC. In reality, CT colonography and OC are likely complementary in that the former tends to assess the right colon particularly well (because of wider lumen and lack of physical constraints), while the latter tends to perform better in the left colon. This suggests that an alternating screening regimen of CT colonography and sigmoidoscopy might be advantageous from clinical, cost-effectiveness, and safety considerations over full colonoscopy alone.

Because our center is among the first to use CT colonography for colorectal screening, it is not surprising that the preferred algorithm for managing discordant polyps has not yet been established. We discovered early in our experience the importance of reviewing these discordant cases because the distinction between a presumed CT colonography FP finding and potential OC FN finding is critical in terms of patient management. To date, our general preference for possible OC FN findings is to first repeat the CT colonography examination and arrange for possible same-day polypectomy if needed. The rationale for this approach is that repeat confirmation with CT colonography bolsters confidence of a true-positive lesion, whereas lack of visualization precludes unnecessary OC. Alternatively, if the test that failed to see the lesion (ie, OC examination) is first repeated and the lesion is still not seen, it may be unclear whether the lesion detected with CT colonography was systematically missed with OC. It is intuitive that the same anatomic factors that led to a missed lesion at OC would remain for the repeat study, even with advanced knowledge, which potentially explains why back-to-back colonoscopy is a flawed design to assess miss rates of OC.

We acknowledge limitations. This study was conducted at a single academic center where we rely on primary physician referrals for our CT colonography program, which may convey some referral bias and limit our ability to generalize the results. CT colonography examinations were performed by using our institutional protocol; differing protocols may affect results. As previously noted, discordant cases without confirmation with subsequent OC were assumed to represent CT colonography FP findings, but it is conceivable that some lesions may have been systematically missed with OC. Finally, additional follow-up of discordant findings was contingent on patient and primary care physician agreement with additional testing; consequently, a small number of patients with discordant findings were lost to follow-up.

In conclusion, we demonstrated that a substantial fraction of discordant lesions found by using CT colonography but not found with subsequent OC represent missed lesions at OC and not CT colonography FP findings. Lesions missed at OC despite a priori knowledge of their size, location, and morphologic structure with CT colonography were, on average, greater than 8 mm in diameter and more likely to be called with higher diagnostic confidence with the initial CT colonography examination. Furthermore, a majority of lesions missed with OC that were ultimately resected demonstrated clinically significant histopathologic results, including a number of advanced premalignant lesions, with a predilection for the right colon, a region where OC is known to have decreased performance. These findings have important implications for continued improvement of colorectal cancer screening strategies.

Advances in Knowledge

• ■ In a clinical practice setting, 21.5% (31 of 144) of discordant polyps 6 mm or greater (ie, detected with CT colonography but not confirmed with subsequent optical colonoscopy [OC]) later proved to be CT colonographic true-positives despite a priori knowledge of CT colonographic findings by endoscopists.

• ■ Of the 78 discordant lesions with subsequent follow-up by OC (and/or CT colonography), 40% (31 of 78) proved to be CT colonography true-positive findings; the remaining 66 discordant polyps without follow-up examinations were assumed to be CT colonographic false-positive findings.

• ■ Of eventually resected false-negative (FN) findings with OC, 81% (21 of 26) were neoplastic (ie, adenomas or serrated lesions) and 43% (nine of 21) were advanced lesions; 89% (eight of nine) of advanced lesions were located in the right colon.

Implications for Patient Care

• ■ Discordant polyps (ie, those detected with CT colonography but not confirmed with subsequent OC) require secondary review and consideration for follow-up because many ultimately prove to be clinically significant lesions.

• ■ The preponderance of advanced right-sided lesions among OC FN findings suggests that CT colonography is useful to evaluate the proximal colon.